Date:         Mon, 1 May 2006 15:43:36 -0400
Reply-To:     "Gerstle, John (CDC/NCHSTP/DHAP) (CTR)" <yzg9@CDC.GOV>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Gerstle, John (CDC/NCHSTP/DHAP) (CTR)" <yzg9@CDC.GOV>
Subject:      Re: cdf plot and distribution fit
Comments: To: Hari Nath <hari_s_nath@YAHOO.COM>
Content-Type: text/plain; charset="us-ascii"

Hari,
Sorry. Thought it did. :)  It's still pretty cool output.

Seems you could create the CDF() from the PDF()/density output.  Don't
know offhand, but you could try:

proc kde data=datawithvariable;
      univar lossvar  / out=lossdensity ;
run;

And then use the resulting dataset, calculate the CDF(), then plot.

John Gerstle, MS
Biostatistician
Northrop Grumman
CDC Information Technological Support Contract (CITS)
NCHSTP \DHAP \HICSB \Research, Analysis, and Evaluation Section
Centers for Disease Control and Prevention

"As far as the laws of mathematics refer to reality, they are not
certain; and as far as they are certain, they do not refer to reality."

                  Albert Einstein, addressing the Prussian Acadamy of
Science, Berlin , Jan 27, 1921

"Boss. We've got cats."  "Meow"

>> -----Original Message-----
>> From: Hari Nath [mailto:hari_s_nath@YAHOO.COM]
>> Sent: Monday, May 01, 2006 3:17 PM
>> To: SAS-L@LISTSERV.UGA.EDU; Gerstle, John (CDC/NCHSTP/DHAP) (CTR)
>> Subject: Re: cdf plot and distribution fit
>>
>> Hi all,
>>     John Gerstle, thanks for your suggestion. I looked up the kde
>> procedure
>> under the univariate and bivariate graphics  - It doesnt produce a
cdf
>> plot, rather it has options for density and histogram for univariate
>> statement
>> --- scatter, contour, histogram and surface plot for the bivariate
>> statement.
>>
>> Does somebody have come across creating a cdf plot.....please let me
>> know...
>>
>> thanks
>> hari
>>
>> On Mon, 1 May 2006 13:12:51 -0400, Gerstle, John (CDC/NCHSTP/DHAP)
(CTR)
>> <yzg9@CDC.GOV> wrote:
>>
>> >Hari,
>> >Look into PROC KDE, using ODS GRAPHICS.
>> >
>> >Something to the effect of:
>> >
>> >goptions   device=gif ;
>> >
>> >ods html file="&outputpath.\Graph of CDF.htm"
>> >        style=statistical gpath="&outputpath.\graphs"
(url="graphs/") ;
>> >ods graphics on / reset;
>> >
>> >ods exclude inputs controls;
>> >proc kde data=datawithvariable;
>> >        univar lossvar  / unistats plots=density;
>> >run;
>> >ods graphics off;
>> >ods html close;
>> >
>> >John Gerstle, MS
>> >Biostatistician
>> >Northrop Grumman
>> >CDC Information Technological Support Contract (CITS)
>> >NCHSTP \DHAP \HICSB \Research, Analysis, and Evaluation Section
>> >Centers for Disease Control and Prevention
>> >
>> >"As far as the laws of mathematics refer to reality, they are not
>> >certain; and as far as they are certain, they do not refer to
reality."
>> >
>> >                  Albert Einstein, addressing the Prussian Acadamy
of
>> >Science, Berlin , Jan 27, 1921
>> >
>> >"Boss. We've got cats."  "Meow"
>> >>> -----Original Message-----
>> >>> From: owner-sas-l@listserv.uga.edu
>> >[mailto:owner-sas-l@listserv.uga.edu]
>> >>> On Behalf Of Hari Nath
>> >>> Sent: Monday, May 01, 2006 1:03 PM
>> >>> To: SAS-L@LISTSERV.UGA.EDU
>> >>> Subject: cdf plot and distribution fit
>> >>>
>> >>> Hi all,
>> >>>    I have a loss variable in the x axis and am trying to create a
cdf
>> >>> plot.
>> >>> Then with that cdf plot, i need to find a best fit for the
available
>> >>> distributions like webull, lognormal, etc.,
>> >>>
>> >>> I do not have the SAS/QC license and Iam not sure how to create a
>> >cdf,
>> >>> apart from using proc capability. But once if i have this cdf
plot
>> >then i
>> >>> could possibly use univariate procedure to get the best fit.
>> >>>
>> >>> Iam using sas 9.1.3 windows.....
>> >>>
>> >>> Any suggestions/ opinions would be helpful.....
>> >>>
>> >>> Thanks in advance,
>> >>> hari